Back to the catalog

ai-gateway

Discover the models and MCP tool servers registered in an AI Gateway and integrate them into any application — call a model over the OpenAI-

Open source Open in the app JSON README (API)

About

Discover the models and MCP tool servers registered in an AI Gateway and integrate them into any application — call a model over the OpenAI-compatible passthrough, connect MCP tools, or scaffold a runnable agent. Read-only against the gateway — consumption only.

Details

Kind
Plugins
Topic
AI, RAG & memory
Publisher
azure
Origin
gemini
Category
ferramentas
Version
2.0.0
Stars
3
Forks
3
Last push
2026-08-27T18:38:26Z
Repository state
ativo
License
MIT
Added
2026-08-30 14:13:39
Updated
2026-08-30 14:13:39
Origin id
azure/ai-gateway

README

# AI Gateway tier in Azure API Management

**One gateway for every model and tool.**
[AI Gateway tier in Azure API Management](https://aka.ms/aigateway/docs) is the AI-focused tier of Azure API
Management. It gives developers one familiar way to call model APIs and MCP tools, while
platform teams enforce security, policy, networking, and observability.

- **Govern model APIs** — Microsoft Foundry, OpenAI, Anthropic, Google Vertex, and more, while developers keep the SDKs they already use.
- **Federate MCP tools** — combine MCP servers, OpenAPI APIs, and SaaS connectors behind one governed MCP endpoint.
- **Enforce AI policies** — access, content safety, token, and rate controls, globally or per model and tool.
- **Observe AI traffic** — token usage, latency, errors, and policy outcomes in dashboards and logs.
- **Run on a trusted platform** — the Azure API Management platform, with Private Link, virtual network integration, and managed identity.

> AI Gateway is in **public preview**, available free of charge. See the **[documentation](https://aka.ms/aigateway/docs)** to get started.

---

## Coding-agent plugin

This repository hosts the official **plugin** for coding agents. It helps
developers **build applications on top of an existing AI Gateway tier in Azure API Management resources**: it discovers the
models and MCP tool servers registered in a gateway and integrates them into whatever
you're building — call a model over the OpenAI-compatible passthrough, connect MCP
tools, or scaffold a runnable agent — all from your coding agent.

The plugin follows the [Claude Code plugin spec](https://code.claude.com/docs/en/plugins),
the [GitHub Copilot / awesome-copilot plugin convention](https://github.com/github/awesome-copilot/blob/main/CONTRIBUTING.md#adding-plugins),
the [Cursor plugin format](https://cursor.com/docs/plugins), and the
[Gemini CLI extension format](https://geminicli.com/docs/extensions/), so it installs
in any of those coding agents (and other agents that support the plugin standard).

> **Consumption only.** The plugin is read-only against the gateway — it never
> provisions, updates, or deletes resources. Creating gateways, models, or tools
> is an administrator task done in the [AI Gateway Portal](https://ai.gateway.azure.com).

### What's inside

| Component | Invoked as | Purpose |
| --------- | ---------- | ------- |
| Skill `use-ai-gateway` | `/ai-gateway:use-ai-gateway` | Full discover → select → credential → integrate workflow (call a model, connect MCP tools, or scaffold an agent). |
| Command `discover` | `/ai-gateway:discover` | Read-only: list a gateway's models and MCP tool servers and pick the ones that fit your use case. |
| Command `build` | `/ai-gateway:build` | End-to-end: discover, retrieve a credential, and integrate the models/tools into your app. |

### Install

The plugin is hosted in this repository, so your coding agent can install it
directly from Git — no manual file downloads. Pick the tab for your agent:

**Claude Code / GitHub Copilot CLI** (plugin marketplace protocol):

```
/plugin marketplace add Azure/ai-gateway
/plugin install ai-gateway@azure
```

**Cursor** — add this repository as a team marketplace from **Dashboard → Plugins →
Add Marketplace → Import from Repo** (`https://github.com/Azure/ai-gateway`), then
install **ai-gateway** from **Customize**. For local testing, symlink the repo into
`~/.cursor/plugins/local/ai-gateway` and reload the window.

**Gemini CLI**:

```
gemini extensions install https://github.com/Azure/ai-gateway
```

### Usage

Point the plugin at your gateway's **ARM resource id** (required to list its models and
tools) and describe what you want to build:

```
/ai-gateway:build /subscriptions/<sub>/resourceGroups/<rg>/providers/Microsoft.ApiManagement/service/<gateway> a service that summarizes support tickets
```

Or explore first, then integrate:

```
/ai-gateway:discover /subscriptions/<sub>/resourceGroups/<rg>/providers/Microsoft.ApiManagement/service/<gateway>
```

### Security

The generated code authenticates the model passthrough and the MCP tool servers
with a gateway **API key**, read from an environment variable
(`AI_GATEWAY_API_KEY`) and never hardcoded. Scaffolded projects keep `.env` out of
source control. If a key is ever exposed, rotate it in the [AI Gateway Portal](https://ai.gateway.azure.com).

To report a security issue, see [SECURITY.md](SECURITY.md).

## Samples

Looking for working examples? The [samples catalog](samples/README.md) curates end-to-end
samples and infrastructure that build on the AI Gateway tier in Azure API Management.

## Contributing

This project welcomes contributions and suggestions. Most contributions require you
to agree to a Contributor License Agreement (CLA) declaring that you have the right
to, and actually do, grant us the rights to use your contribution. For details,
visit [Contributor License Agreements](https://cla.opensource.microsoft.com).

When you submit a pull request, a CLA bot will automatically determine whether you
need to provide a CLA and decorate the PR appropriately (e.g., status check,
comment). Simply follow the instructions provided by the bot. You will only need to
do this once across all repos using our CLA.

This project has adopted the [Microsoft Open Source Code of Conduct](https://opensource.microsoft.com/codeofconduct/).
For more information see the [Code of Conduct FAQ](https://opensource.microsoft.com/codeofconduct/faq/)
or contact [opencode@microsoft.com](mailto:opencode@microsoft.com) with any
additional questions or comments.

## Trademarks

This project may contain trademarks or logos for projects, products, or services.
Authorized use of Microsoft trademarks or logos is subject to and must follow
[Microsoft's Trademark & Brand Guidelines](https://www.microsoft.com/legal/intellectualproperty/trademarks/usage/general).
Use of Microsoft trademarks or logos in modified versions of this project must not
cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or
logos are subject to those third-party's policies.

More